| Issue Number | 4089 |
|---|---|
| Summary | Improve/document how versions of filters and control documents are used by the publishing system |
| Created | 2016-05-04 11:58:08 |
| Issue Type | Improvement |
| Submitted By | Kline, Bob (NIH/NCI) [C] |
| Assigned To | Kline, Bob (NIH/NCI) [C] |
| Status | Closed |
| Resolved | 2018-03-14 12:41:16 |
| Resolution | Fixed |
| Path | /home/bkline/backups/jira/ocecdr/issue.183608 |
During deployment and testing of Darwin we discovered that the web
front end for submitting a publishing job and the software which
processes the publishing job use different approaches for fetching the
publishing control document. The web front end fetches the most recent
publishable version of the document (even though the control document
itself isn't being published), whereas the publishing software itself
gets the latest version of the control document, regardless of the value
in the publishable column of the document version table.
The script to verify that the repository has control documents which
match the current version control branch was using a third approach,
comparing the version control file with the current working document,
even if it contained unversioned changes. That script has been modified
to match the web front end, at least temporarily, using the latest
publishable version of each control document for the check against the
version control branch. Please make the publishing software and the web
front end use the same approach for retrieving publishing control
documents, and ensure that Publishing/check-control-docs.py
uses the same approach, too.
In addition, please determine how the publishing software figures out
which version of the filters to use, and make sure that
Filters/compare-filters.py uses the same approach.
Story point estimate may be overly conservative (that is, high), but this is in an important and complex part of the system, and we want to be especially careful not to break anything.
~volker: I propose that we
use the most recent version of the publishing control document and the
filter documents, regardless of the publishable column's
value. What do you think?
Yes, I think that makes a lot of sense. I can't tell you how often I
ran a test publishing job - maybe even over night - just to find out at
the end of the job that there were no changes to the output because I
had forgotten to save a filter as a publishable document.
Since the filters and control documents aren't published, i.e.
exported to our partners, there is no need to having to store these as
publishable versions.
The publishing software now ignores the publishable flag
for filters and the publishing control doc. The publishing job itself
uses the latest versions of filters and the publishing control document
created before the job started. Installed on DEV.
https://github.com/NCIOCPL/cdr-publishing/commit/449b932
https://github.com/NCIOCPL/cdr-publishing/commit/a17ec24
https://github.com/NCIOCPL/cdr-admin/commit/307a73d7
https://github.com/NCIOCPL/cdr-lib/commit/ae5ed63
Elapsed: 0:00:00.001368